Show HN: Agent-relay for multi-vendor coding sessions (github.com)

A new tool called agent-relay has been introduced to facilitate collaboration between AI coding sessions from different vendors, such as Claude Code and Codex. This prototype allows these distinct AI systems to communicate seamlessly by managing context transfer automatically, significantly reducing the manual task of contextual handover between coding sessions. The agent-relay acts as a relay where one session leads and another follows, storing records and pausing for human decisions when needed. This streamlines collaborative coding processes, enabling efficient planning, building, and reviewing of code. The significance of agent-relay lies in its ability to improve productivity for developers working with multiple AI coding agents. By allowing AI models to review each other's work and maintaining a transparent audit trail of decisions and messages, developers can ensure higher quality outputs without constant oversight. Key features include a web console for monitoring progress, the ability to set job parameters like turn limits, and extensibility through an adapter system, which opens up opportunities for integration with various coding agents. This innovation presents a pathway for more sophisticated agent-to-agent interactions within the AI/ML community, paving the way for more coordinated and less resource-intensive AI collaborations.